Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,180 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
task_categories:
|
| 4 |
+
- question-answering
|
| 5 |
+
language:
|
| 6 |
+
- en
|
| 7 |
+
tags:
|
| 8 |
+
- finance
|
| 9 |
+
- music
|
| 10 |
+
- medical
|
| 11 |
+
- food
|
| 12 |
+
- academic disciplines
|
| 13 |
+
- natural disasters
|
| 14 |
+
- software
|
| 15 |
+
- synthetic
|
| 16 |
+
pretty_name: Using KGs to test knowledge consistency in LLMs
|
| 17 |
+
size_categories:
|
| 18 |
+
- 10K<n<100K
|
| 19 |
+
---
|
| 20 |
+
|
| 21 |
+
For background on this dataset, please check https://arxiv.org/abs/2405.20163.
|
| 22 |
+
|
| 23 |
+
## What it is:
|
| 24 |
+
Each dataset in this delivery is made up of query clusters that test an aspect of the consistency of the LLM knowledge about a particular domain. All the questions in each
|
| 25 |
+
cluster are meant to be answered either 'yes' or 'no'. When the answers vary within a cluster, the knowledge is said to be inconsistent. When all the questions in a cluster
|
| 26 |
+
are answered 'no' when the expected answer is 'yes' (or viceversa), the knowledge is said to be 'incomplete' (i.e., maybe the LLM wasn't trained in that particular domain).
|
| 27 |
+
It is our experience that incomplete clusters are very few (less than 3%) meaning that the LLMs we have tested know about the domains included here (see below for a list of the
|
| 28 |
+
individual datasets), as opposed to inconsistent clusters, which can be between 6%-20% of the total clusters.
|
| 29 |
+
|
| 30 |
+
## How it is made:
|
| 31 |
+
The questions and clusters are automatically generated from a knowledge graph from seed concepts and properties. In our case, we have used Wikidata,
|
| 32 |
+
a well known knowledge graph. The result is an RDF/OWL subgraph that can be queried and reasoned over using Semantic Web technology.
|
| 33 |
+
|
| 34 |
+
## Types of query clusters
|
| 35 |
+
|
| 36 |
+
There are different types of query clusters depending on what aspect of the knowledge graph and its deductive closure they capture:
|
| 37 |
+
|
| 38 |
+
Edge clusters test a single edge using different questions. For example, to test the edge ('orthopedic pediatric surgeon', IsA, 'orthopedic surgeon), the positive
|
| 39 |
+
or 'edge_yes' (expected answer is 'yes') cluster is:
|
| 40 |
+
|
| 41 |
+
"is 'orthopedic pediatric surgeon' a subconcept of 'orthopedic surgeon' ?",
|
| 42 |
+
"is 'orthopedic pediatric surgeon' a type of 'orthopedic surgeon' ?",
|
| 43 |
+
"is every kind of 'orthopedic pediatric surgeon' also a kind of 'orthopedic surgeon' ?",
|
| 44 |
+
"is 'orthopedic pediatric surgeon' a subcategory of 'orthopedic surgeon' ?"
|
| 45 |
+
|
| 46 |
+
There are also inverse edge clusters (with questions like "is 'orthopedic surgeon' a subconcept of 'orthopedic pediatric surgeon' ?") and negative or 'edge_no' clusters
|
| 47 |
+
(with questions like "is 'orthopedic pediatric surgeon' a subconcept of 'dermatologist' ?")
|
| 48 |
+
|
| 49 |
+
Hierarchy clusters measure the consistency of a given path, including n-hop virtual edges (in graph's the deductive closure). For example, the path
|
| 50 |
+
('orthopedic surgeon', 'surgeon', 'medical specialist', 'medical occupation') is tested by the cluster below
|
| 51 |
+
|
| 52 |
+
"is 'orthopedic surgeon' a subconcept of 'surgeon' ?",
|
| 53 |
+
"is 'orthopedic surgeon' a type of 'surgeon' ?",
|
| 54 |
+
"is every kind of 'orthopedic surgeon' also a kind of 'surgeon' ?",
|
| 55 |
+
"is 'orthopedic surgeon' a subcategory of 'surgeon' ?",
|
| 56 |
+
"is 'orthopedic surgeon' a subconcept of 'medical specialist' ?",
|
| 57 |
+
"is 'orthopedic surgeon' a type of 'medical specialist' ?",
|
| 58 |
+
"is every kind of 'orthopedic surgeon' also a kind of 'medical specialist' ?",
|
| 59 |
+
"is 'orthopedic surgeon' a subcategory of 'medical specialist' ?",
|
| 60 |
+
"is 'orthopedic surgeon' a subconcept of 'medical_occupation' ?",
|
| 61 |
+
"is 'orthopedic surgeon' a type of 'medical_occupation' ?",
|
| 62 |
+
"is every kind of 'orthopedic surgeon' also a kind of 'medical_occupation' ?",
|
| 63 |
+
"is 'orthopedic surgeon' a subcategory of 'medical_occupation' ?"
|
| 64 |
+
|
| 65 |
+
Property inheritance clusters test the most basic property of conceptualization. If an orthopedic surgeon is a type of surgeon, we expect that
|
| 66 |
+
all the properties of surgeons, e.g., having to be board certified, having attended medical school or working on the field of surgery, are inherited by orthopedic surgeons.
|
| 67 |
+
The example below tests the later:
|
| 68 |
+
|
| 69 |
+
"is 'orthopedic surgeon' a subconcept of 'surgeon' ?",
|
| 70 |
+
"is 'orthopedic surgeon' a type of 'surgeon' ?",
|
| 71 |
+
"is every kind of 'orthopedic surgeon' also a kind of 'surgeon' ?",
|
| 72 |
+
"is 'orthopedic surgeon' a subcategory of 'surgeon' ?",
|
| 73 |
+
"is the following statement true? 'orthopedic surgeon works on the field of surgery' ",
|
| 74 |
+
"is the following statement true? 'surgeon works on the field of surgery' ",
|
| 75 |
+
"is it accurate to say that 'orthopedic surgeon works on the field of surgery'? ",
|
| 76 |
+
"is it accurate to say that 'surgeon works on the field of surgery'? "
|
| 77 |
+
|
| 78 |
+
|
| 79 |
+
|
| 80 |
+
## List of datasets
|
| 81 |
+
|
| 82 |
+
To show the versatility of our approach, we have constructed similar datasets in the domains below. We test one property inheritance per dataset. The Wikidata main QNode
|
| 83 |
+
(the node corresponding to the entities) and PNode (the node corresponding to the property) are indicated in parenthesis.
|
| 84 |
+
|
| 85 |
+
### ACADEMIC_DISCIPLINES (https://www.wikidata.org/wiki/Q11862829) ONTOLOGY -- V1 = 443 CLUSTERS, "has use" (https://www.wikidata.org/wiki/Property:P366)
|
| 86 |
+
|
| 87 |
+
edges_yes = 52
|
| 88 |
+
|
| 89 |
+
edges_no = 308
|
| 90 |
+
|
| 91 |
+
edges_inv = 52
|
| 92 |
+
|
| 93 |
+
hierarchies = 30
|
| 94 |
+
|
| 95 |
+
property hierarchies = 1
|
| 96 |
+
|
| 97 |
+
### DISHES (https://www.wikidata.org/wiki/Q746549) ONTOLOGY -- V1 = 1220 CLUSTERS, has parts (https://www.wikidata.org/wiki/Property:P527) --> has ingredient
|
| 98 |
+
|
| 99 |
+
edges_yes = 225
|
| 100 |
+
|
| 101 |
+
edges_no = 521
|
| 102 |
+
|
| 103 |
+
edges_inv = 224
|
| 104 |
+
|
| 105 |
+
hierarchies = 72
|
| 106 |
+
|
| 107 |
+
property hierarchies = 178
|
| 108 |
+
|
| 109 |
+
### FINANCIAL PRODUCT (https://www.wikidata.org/wiki/Q15809678) ONTOLOGY -- V1: 725 CLUSTERS, "used by" (https://www.wikidata.org/wiki/Property:P1535)
|
| 110 |
+
|
| 111 |
+
edges_yes = 112
|
| 112 |
+
|
| 113 |
+
edges_no = 433
|
| 114 |
+
|
| 115 |
+
edges_inv = 108
|
| 116 |
+
|
| 117 |
+
hierarchies = 40
|
| 118 |
+
|
| 119 |
+
property hierarchies = 32
|
| 120 |
+
|
| 121 |
+
### HOME APPLIANCES (https://www.wikidata.org/wiki/Q212920) ONTOLOGY -- V1 = 421 CLUSTERS, "has use" (https://www.wikidata.org/wiki/Property:P366)
|
| 122 |
+
|
| 123 |
+
edges_yes = 58
|
| 124 |
+
|
| 125 |
+
edges_no = 261
|
| 126 |
+
|
| 127 |
+
edges_inv = 58
|
| 128 |
+
|
| 129 |
+
hierarchies = 31
|
| 130 |
+
|
| 131 |
+
property hierarchies = 13
|
| 132 |
+
|
| 133 |
+
### MEDICAL SPECIALTIES (https://www.wikidata.org/wiki/Q930752) ONTOLOGY -- V1 = 740 CLUSTERS, "field of occupation" (https://www.wikidata.org/wiki/Property:P425)
|
| 134 |
+
|
| 135 |
+
edges_yes = 122
|
| 136 |
+
|
| 137 |
+
edges_no = 386
|
| 138 |
+
|
| 139 |
+
edges_inv = 114
|
| 140 |
+
|
| 141 |
+
hierarchies = 55
|
| 142 |
+
|
| 143 |
+
property hierarchies = 63
|
| 144 |
+
|
| 145 |
+
### MUSIC_GENRES (https://www.wikidata.org/wiki/Q188451) ONTOLOGY -- V1 = 1990 CLUSTERS, "practiced by" (https://www.wikidata.org/wiki/Property:P3095)
|
| 146 |
+
|
| 147 |
+
edges_yes = 490
|
| 148 |
+
|
| 149 |
+
edges_no = 807
|
| 150 |
+
|
| 151 |
+
edges_inv = 488
|
| 152 |
+
|
| 153 |
+
hierarchies = 212
|
| 154 |
+
|
| 155 |
+
property hierarchies = 139
|
| 156 |
+
|
| 157 |
+
|
| 158 |
+
### NATURAL DISASTERS (https://www.wikidata.org/wiki/Q8065) ONTOLOGY -- V1 = 357 CLUSTERS, "has cause" (https://www.wikidata.org/wiki/Property:P828)
|
| 159 |
+
|
| 160 |
+
edges_yes = 45
|
| 161 |
+
|
| 162 |
+
edges_no = 225
|
| 163 |
+
|
| 164 |
+
edges_inv = 44
|
| 165 |
+
|
| 166 |
+
hierarchies = 21
|
| 167 |
+
|
| 168 |
+
property hierarchies = 22
|
| 169 |
+
|
| 170 |
+
### SOFTWARE (https://www.wikidata.org/wiki/Q7397) ONTOLOGY -- V1: 849 CLUSTERS, "studied in" (https://www.wikidata.org/wiki/Property:P7397) is the property
|
| 171 |
+
|
| 172 |
+
edges_yes = 80
|
| 173 |
+
|
| 174 |
+
edges_no = 572
|
| 175 |
+
|
| 176 |
+
edges_inv = 79
|
| 177 |
+
|
| 178 |
+
hierarchies = 114
|
| 179 |
+
|
| 180 |
+
property hierarchies = 4
|